Early high protein intake is associated with low mortality and energy overfeeding with high mortality in non-septic mechanically ventilated critically ill patients

Picture a patient on day four of mechanical ventilation. They can't eat, can't speak, and can't tell anyone what they need. A tube delivers everything — fluids, sedation, and nutrition. A physician and a nutritionist are standing at the bedside, arguing over how much protein to run through that tube. Both believe they're helping. The tragedy is that one of them might be wrong in a way that kills the patient. Weijs and colleagues set out to find which decisions in that window actually mattered, and the answer upended some basic instincts about feeding the critically ill. The background to this study is genuine scientific conflict. Clinical guidelines recommend protein targets between 1.2 and 1.5 grams per kilogram of body weight per day, with some expert opinion pushing even higher. However, no randomized trial had specifically tested early protein feeding in the intensive care unit. The observational data pointed in different directions. A post-hoc analysis of the EPaNIC trial linked early protein and amino acid provision to delayed recovery. A small study found that changes in muscle mass during the first week and a half of an ICU stay were associated with protein intake. Post-mortem muscle biopsies in 12 patients showed impaired autophagy — the cellular housekeeping process that removes damaged proteins and organelles — correlated with the amount of amino acids infused. The field was stuck between the intuition that protein protects muscle and the possibility that early protein might, under certain conditions, do harm. Energy prescription added a second layer of uncertainty. Several trials suggested that early low-energy, or trophic, feeding did not worsen survival. But in many of those trials, energy targets were estimated, not measured. That distinction turns out to matter enormously. The study Weijs and colleagues built was designed to cut through this noise. Between August two thousand four and March two thousand ten, four thousand eight hundred and three patients were admitted to a mixed medical-surgical university intensive care unit in the Netherlands. After applying inclusion criteria — mechanically ventilated for more than 72 hours, predicted to need artificial nutrition for at least another five to seven days, and first intensive care unit admission — eight hundred and forty-three patients were analyzed. The key methodological choice was that energy expenditure was measured, not guessed. Indirect calorimetry using a Deltatrac metabolic monitor captured actual energy burn while enteral feeding continued, giving the researchers a real denominator rather than an estimate. Energy overfeeding was then defined precisely: energy intake exceeding one hundred ten percent of measured expenditure by day four. Protein intake was divided into four bands using day-four values adjusted for body mass index: below zero point eight, zero point eight to one point zero, one point zero to one point two, and one point two grams per kilogram or above. Logistic regression analyzed the relationship between these nutritional variables, sepsis status, and hospital mortality, adjusting throughout for Acute Physiology and Chronic Health Evaluation II — or APACHE II — score, which is the standard illness-severity index in critical care. The analysis was built to find a three-way interaction: protein level, overfeeding, and whether the patient had sepsis. The headline finding lands hard. Among non-septic patients who were not overfed — four hundred and nineteen people — hospital mortality fell in a near-linear staircase as day-four protein intake increased. At below zero point eight grams per kilogram, mortality was thirty-seven percent. At zero point eight to one point zero grams per kilogram, it was thirty-five percent. At one point zero to one point two grams per kilogram, it dropped to twenty-six point five percent. At one point two grams per kilogram and above, it fell to nineteen percent — a statistically significant gradient with a p-value of zero point zero three three. When the team split the group at the one point two grams per kilogram threshold, patients above it had an adjusted odds ratio for death of zero point four two. That means roughly half the odds of dying compared with patients below that threshold, even after accounting for illness severity. The difference in absolute protein between the worst and best groups is not enormous in clinical terms. But in this dataset, it was the difference between a one in three and a one in five chance of dying in the hospital. Then comes the critical carve-out that makes the finding scientifically coherent rather than just a number. In patients admitted with sepsis, the protein signal disappeared entirely. The one hundred seventeen septic patients had an overall higher mortality — about forty-nine percent versus thirty-four percent in non-septic patients. Logistic regression in that group showed no relationship whatsoever between day-four protein intake and mortality: an odds ratio of one point one five, with a confidence interval running from zero point eight to one point six six, and a p-value of zero point four six. The authors offer a biological rationale rooted in autophagy. Autophagy doesn't just clear damaged cellular debris; it also degrades intracellular microorganisms, which means it's part of the immune response to bacterial infection. Protein intake suppresses autophagy. In a septic patient, blunting that cellular self-defense mechanism at the moment of peak bacterial challenge might remove something the body actually needs. In non-septic critical illness, that same suppression may be less consequential, leaving the muscle-protective effects of protein to dominate. The sepsis carve-out isn't a statistical anomaly — it's biology working as predicted. Now pivot to the second major finding: overfeeding kills. In the full cohort, patients who received more than one hundred ten percent of their measured energy needs by day four had an adjusted odds ratio for hospital mortality of one point six two — a sixty-two percent higher chance of dying compared with patients who were not overfed, after adjustment for APACHE II and other variables. In the non-septic subgroup alone, that odds ratio rose to one point eight nine. And the prevalence of overfeeding was striking: forty-one percent of all patients, and three hundred seven of seven hundred twenty-six non-septic patients, were classified as overfed on day four. Nearly half. The paper traces this directly to standard clinical practice. The Harris-Benedict equation plus a thirty percent activity and stress addition is a common shortcut for estimating intensive care unit energy needs, and according to their data, this approach is an inaccurate predictor in intensive care unit patients — associated with significant overfeeding. Early in critical illness, endogenous glucose production may supply more than half of energy expenditure, meaning the body is already generating substantial fuel internally. Pouring in full external energy on top of that creates a metabolic mismatch. The consequences sketched in the paper include excess infection, prolonged mechanical ventilation, and impaired autophagy. More food in the intensive care unit is not safer. Measured targets are. These two findings interlock. The protein benefit only emerged in patients who were not being overfed. When you overfeed on energy, the protein signal is masked — or perhaps actively reversed. Weijs and colleagues found that in overfed patients, day-four protein intake was not associated with mortality in either direction. The implication is that getting the energy right is the prerequisite. Only when energy is appropriately calibrated does the protein level become a meaningful lever. These two variables cannot be treated independently at the bedside — they interact, and the interaction has consequences. One more piece of the architecture deserves acknowledgment. Energy overfeeding was itself an independent predictor of mortality alongside sepsis status. Sepsis carried an adjusted odds ratio of one point seven seven in the full cohort. Overfeeding came in at one point six two. These are comparable in magnitude. In a field where sepsis is understood as one of the most dangerous conditions in critical care, finding that a nutritional practice carries a mortality signal of similar size should carry weight. The study's limitations are real, and the authors state them clearly. This is a post-hoc observational analysis. Protein intake reflects what patients happened to receive, not a randomized assignment. Lower protein intake in some patients might reflect greater illness severity in ways the APACHE II score doesn't fully capture. Indirect calorimetry was often performed after day four, meaning overfeeding may have continued for several days before targets were corrected. The authors call explicitly for a randomized controlled trial to confirm the benefit of early high-protein feeding in non-septic patients before it can be adopted as a recommendation. But the data are real, and the cohort is substantial. Eight hundred and forty-three mechanically ventilated patients, energy expenditure measured by indirect calorimetry while feeding continued, a prospective database spanning nearly six years. The signal is consistent, the subgroup finding is biologically grounded, and the magnitude is clinically meaningful. Reaching or exceeding one point two grams per kilogram of protein by day four, while keeping energy intake calibrated to actual measured expenditure rather than estimated formulas, was associated with roughly half the mortality risk compared with lower protein in non-septic patients. For a person on a ventilator who cannot speak for themselves, that gap is everything. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.